Infornation Extraction from Financial News for Market Indices Forecast

This work is about market rates forecast based on information extraction based on news texts. Firstly, in the thesis there was mentioned some theoretical aspects on this topic. Next, the already done researches in this field were analyzed. Finally, the author’s own research implementation was described. For this scope the corpus consisting of Microsoft and Apple companies was created (by crawling them from the internet). The Subject – Predicate – Object triplets were extracted from these texts. After that the triplets were transformed to vectors using one-hot method, which became the input data for an artificial neural network. The obtained results are quite comparable with the obtained ones from other researchers. Also in the work there was mentioned the ways of improvement the implemented algorithm. First of all, the research would be useful for the companies in risks analysis and strategies implementation. Also it would be interesting for the people who works under market rate predictions